ChatGPT Prompt to Write a Cold LinkedIn DM That Gets Replies

A founder selling a member-churn early-warning tool needed one meeting with a gym chain to prove the model worked. She had a list of forty operations directors, a demo she believed in, and a reply rate of roughly two percent from the copy-paste template her previous contractor left behind. The tool predicted which members were about to cancel based on a drop in check-ins over thirty days, and for a chain bleeding nine percent of its members a quarter, that was a real number, not a gimmick. The messages still died in the request folder, because every one of them opened with “I hope this finds you well” and closed with “let me know if you’d like to hop on a quick call.”

The prompt below writes a cold LinkedIn DM that leads with the buyer’s specific pain, earns the right to a two-line ask, and stays short enough to survive a phone screen. It is built for pitching that churn tool to a multi-location gym operator, and the fields you swap let it fit any B2B cold outreach where the trigger is a number the buyer already feels.

The Scenario Behind the Prompt

Picture Ironpath Fitness, a chain with twenty-two clubs and a monthly cancel rate that has crept up for three straight quarters. The buyer is the operations director, whose bonus is tied to retention and who already suspects that a large share of cancellations were decided weeks before the member actually quit. She does not need to be told that retention matters. She needs a reason to believe that a specific, small signal, a drop in check-ins, can be caught early enough to act on. That belief is the entire job of the first message.

Cold DMs fail for a predictable reason. They open with the sender, describe the product, and ask for time before they have given the reader a single useful idea. The prompt below reverses that order: a fact about the buyer’s world, one sentence of how the tool changes it, and a low-friction question rather than a meeting request.

The Copy-Paste LinkedIn DM Prompt

Paste the whole block into ChatGPT and replace the fields at the top. It writes three messages: the first cold touch, a short follow-up, and a final close-the-loop note.

You are a B2B sales copywriter who writes short, credible cold LinkedIn messages.

Write a three-message cold outreach sequence to a specific buyer.

Sender: {{SENDER}} sells {{PRODUCT}}, which {{ONE_LINE_MECHANISM}}.
Target buyer: {{BUYER_ROLE}} at {{COMPANY_TYPE}}.
Trigger or pain the buyer already feels: {{TRIGGER}}.
One proof point I can honestly claim: {{PROOF}}.
Desired next step: {{ASK}}.
What I must never say: {{BANNED}}.

Rules:
1. Message 1 opens with the buyer's problem or a sharp fact about their world, never with my name or my company.
2. Keep message 1 under 90 words. No greeting longer than two words.
3. Name the mechanism in one sentence, in plain language, with no buzzwords.
4. Include exactly one proof point, and only if it is factual. Do not invent statistics.
5. End message 1 with a soft, specific question, not a meeting request.
6. Message 2 is a two-sentence follow-up sent four days later, adds one new angle, and does not repeat message 1.
7. Message 3 is a polite close-the-loop note, three sentences maximum, that gives the buyer an easy exit.
8. Use short sentences and a conversational tone. No exclamation marks, no emojis, no "I hope this finds you well."
9. After each message, add one line explaining what it is trying to earn.

The rule that message 1 must never open with the sender is the load-bearing one. A reader decides in under three seconds whether a message is about them or about you, and most cold DMs lose that second before the product is even named.

Fields to Set Before You Send

Fill {{PROOF}} with something you can defend. A single verifiable pilot result converts better than a vague claim about “thousands of gyms,” and a fabricated number is the fastest way to lose a professional buyer who will check. If you have no proof yet, say so in the field and the model will write around the gap instead of inventing one.

What the Model Should Hand Back

Expect three short messages with a rationale line under each. A trimmed slice of the expected shape:

Message 1 - subject not needed, keep it under 90 words

Most gym cancellations are decided about three weeks before the member actually quits.
The signal is usually a drop in check-ins that nobody is watching in time.

[One sentence on how the dashboard flags a 40 percent check-in drop over 30 days.]

[One factual pilot result.]

Out of curiosity, how far in advance does your team usually see a cancellation coming?

Job: replaces a pitch with a question the buyer cannot answer instantly, which earns a reply.

Message 2 - four days later
[One new angle, such as the cost of a single saved member versus the tool.]
...

Message 3 - close the loop
[Three sentences, easy exit, no pressure.]
...

A good first message should feel like a useful observation from someone who knows the industry, not a sales hook. If all three messages restate the same pitch, you have a sequence that is technically complete but reads as nagging, and you should push message 2 to carry genuine new information, such as what a saved member is worth against the subscription cost.

Variations: Warmer, Shorter, Different Buyer

• Warm lead. If the buyer viewed your profile or opened a previous message, add “Acknowledge that they engaged but do not say you noticed; reference the topic they care about.”
• Mobile-first version. Ask for “a single message under 60 words with no line breaks longer than one sentence” for buyers who read on a phone between meetings.
• Different stakeholder. Retarget the same prompt at the Head of Member Experience by changing {{TRIGGER}} to the pain that role owns, such as on-boarding drop-off in the first 90 days.
• Referral angle. Add “Open with a shared connection’s name and why they thought this was relevant.”
• Region and tone. Add “Use a direct, low-context tone suitable for a US buyer” or a warmer register for a different market, since cold outreach reads very differently across regions.

Send the sequence, track reply rate per message, and rewrite the message with the lowest reply. For the adjacent craft of cold email that goes long, read the cold email prompt. To adapt the same voice to a public feed, see ChatGPT prompts for LinkedIn, and to sharpen who you are actually writing for, pair it with the B2B and B2C AI persona guide.

Why This Beats a Template Blast

Three choices carry the result. First, leading with the buyer’s problem earns the three seconds a template loses with a greeting. Second, naming the mechanism in one plain sentence proves you understand the tool’s own logic, which is what separates a thoughtful pitch from a mass send. Third, ending with a question rather than a meeting request lowers the cost of replying to almost nothing. Together they produce a message that reads like a person who did ten minutes of homework, which at forty companies a week is enough to move a reply rate from two percent to something a founder can build a pipeline on.

Sources & Notes

This prompt is written for ChatGPT (GPT-4o and later) and works in Claude and Gemini with minor rewording. The observed behavior that opening with the buyer’s problem outperforms an opener about the sender reflects general cold-outreach practice as of 2024-2025, not a controlled study, so test on your own list. No external data or benchmarks were fetched. The gym chain, buyer, product, and pilot figures here are illustrative placeholders. Before you send anything, confirm every proof point is accurate and compliant with LinkedIn’s messaging rules and your local outreach regulations, and respect contact-frequency limits.